© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Andy Arditi · Finding Misaligned Persona Features in Open-Weight Models
    5. Llama3.1-8B-IT
    7. Resid Post - 131k
    9. 11-RESID-POST-AA
    11. 13916
    Prev
    Next
    INDEX
    Explanations

    problems, solutions, processes

    np_max-act · gemini-2.0-flash

    This neuron activates on instruction-related keywords—especially “trend/trends” and explanation cues like “why,” “how,” “should,” and “could.”

    oai_token-act-pair · o4-miniTriggered by @xinyanhu8

    The neuron detects words and phrases expressing possibility, contingency, or speculation (modal/conditional language like could, would, might, think, when).

    oai_token-act-pair · gpt-5-miniTriggered by @vetterc0
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    andyrdt/saes-llama-3.1-8b-instruct/resid_post_layer_11/trainer_1
    Dataset (Dashboard)
    Various
    Features
    131,072
    Data Type
    float32
    Hook Name
    blocks.11.hook_resid_post
    Architecture
    standard
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
     damaged
    -0.07
    ennial
    -0.07
    Remote
    -0.07
    trie
    -0.07
    fitness
    -0.07
    arket
    -0.07
     guarantees
    -0.06
    ù
    -0.06
    owns
    -0.06
    .ph
    -0.06
    POSITIVE LOGITS
     grupo
    0.07
     chemicals
    0.06
     цих
    0.06
    -ignore
    0.06
    _frac
    0.06
    \-
    0.06
     мы
    0.06
    \HttpFoundation
    0.06
    丁
    0.06
     اینکه
    0.06
    Activations Density 0.125%

    No Known Activations